Asymmetric Server Partitioning for Cluster Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current server clustering techniques require all nodes to host the same set of projects, leading to overhead costs and performance issues, especially during project maintenance, where unexpected errors can cause downtime and affect other projects.
Innovation Solution
Implementing an asymmetric project partitioning scheme where each server can host a unique subset of projects, using project partition modules to determine and manage project distribution based on user preferences, weighted balancing rules, and redundancy rules, allowing for dynamic re-partitioning and load balancing across the cluster.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all nodes in a cluster host the exact same set of projects, then system reliability is improved through redundancy, but overhead costs increase and scalability deteriorates
Solution Approach 1:
The patent segments the cluster into different node types (master nodes and worker nodes) with distinct project hosting responsibilities. Master nodes host control plane projects while worker nodes host data plane projects, allowing each node to host only the projects necessary for its function rather than all projects across the cluster. This segmentation reduces redundant overhead while maintaining reliability through specialized distribution.
2Reliability
If all nodes in a cluster host the exact same set of projects, then fault tolerance is improved, but adding new projects negatively affects overall cluster performance
Solution Approach 1:
The patent applies local quality by assigning different project sets to different nodes based on their roles and capabilities. Master nodes maintain control plane projects with high availability requirements, while worker nodes host data plane projects that can be dynamically added or removed. This allows new projects to be added to worker nodes without impacting the stability and performance of control plane projects on master nodes.
3Adaptability or versatility
If projects are distributed across multiple servers, then maintenance flexibility is improved, but unexpected errors can still cause server restarts affecting other projects
Solution Approach 1:
The patent segments projects into control plane projects on master nodes and data plane projects on worker nodes. This segmentation isolates failures: when a worker node experiences an error requiring restart, only data plane projects on that specific worker are affected, while control plane projects on master nodes continue operating. This maintains both maintenance flexibility and project availability through architectural isolation.
Data Source
AI summary
A technique for project partitioning in a cluster of servers is disclosed. In one particular exemplary embodiment, the technique may be realized as a method for partitioning a plurality of projects among a cluster of servers. The method may comprise determining a status of project partition among the cluster of servers. The method may also comprise generating a partition plan for re-partitioning the plurality of projects among the cluster of servers based at least in part on the status and a partition policy, wherein the partition policy comprises one or more rules selected from a group consisting of a user specified project-server association rule, a weighted balancing rule, and a project redundancy rule. The method may further comprise communicating the partition plan to each of the cluster of servers. The method may additionally comprise causing the plurality of projects to be re-partitioned among the cluster of servers based on the partition plan.